基于多普勒量测的UKF多目标跟踪方法

UKF Multitarget Tracing with Doppler Information

针对高斯混合概率假设密度滤波(GMPHD)初始化需要先验强度函数的缺点,在现有GMPHD滤波框架的基础上,将多普勒信息附加于滤波更新过程中,提出了基于多普勒量测的UKF多目标跟踪方法。该方法能够利用多普勒信息对新生目标状态初始化,实现新生目标的自动初始化,有效降低滤波估计误差。仿真结果表明,所提方法复杂度较低,且在多目标跟踪过程中,对于目标个数的估计精度和最优子模式指派距离均优于已有方法。

The standard Gaussian mixture probability hypothesis density(GMPHD)filter is a promising algorithm for multi-target tracking.However,the performance of the algorithm is greatly influenced by clutters and the initialization is restricted by a priori information.To solve this problem,on the basis of the GMPHD prediction,Doppler information was appended to the filtering process to improve the performance of the algorithm.Firstly,a concrete initialization process was proposed in the birth intensity design of the GMPHD.The initialization process from consecutive measurements led to a reliable birth intensity that improved track management performance.Secondly,in order to measure the nonlinearity,the unscented Kalman filter(UKF)was used to replace the Kalman filter to filter the measurement vector.Simulation results showed that the proposed algorithm improve the accuracy of target number estimation was as well as the optimal subpattern assignment distance when compared with the existing algorithm.